Method for categorizing digital video data

Abstract

In one aspect, a computer implemented method for extracting information on actions or activities captured in digital data includes steps of: providing a first digital data set, extracting and characterizing first features from the first digital data set, creating first clusters from the extracted first features according to probability density functions and reference function distance calculations, providing a second digital data set, extracting and characterizing second features from the second digital data set, characterizing the extracted second features according to probability density functions and reference function distance calculations, and matching the second extracted features to a portion of the first extracted features clusters according to similarities in reference function distance calculations.

Description

FIELD OF THE INVENTION[0001]The invention relates to methods for categorizing digital data. The invention relates particularly to categorizing digital video data by tagging actions captured in the data.BACKGROUND OF THE INVENTION[0002]The automated recognition of objects and elements in images is known in the art. Faces, smiles people, animals may each be detected in images using automated computer analysis systems. Recognition of actions captured in images remains a challenge as actions may require abstract conceptualization, are not easily captured in still images, defining the target action to be tagged can be nuanced, for example: walking versus limping versus running as the action to be tagged. Typical methods for utilizing artificial intelligence systems including deep learning systems, require providing training data sets with both input data and identified output results. In the case of image tagging such a system may require extensive manual intervention to construct a trai...

AI Technical Summary

Benefits of technology

The patent describes a computer method for extracting information about actions or activities captured in digital data. The method involves analyzing data sets, creating clusters of features based on probability density functions and reference function distance calculations, and matching features from different data sets based on their similarity. The technical effect of this patent is an improved ability to extract information about actions and activities from digital data.

Problems solved by technology

Recognition of actions captured in images remains a challenge as actions may require abstract conceptualization, are not easily captured in still images, defining the target action to be tagged can be nuanced, for example: walking versus limping versus running as the action to be tagged.

Each training set may be of limited utility or tagging range.

Each activity may require the construction of an individual data set requiring cost prohibitive amounts of time, and resources, including human supervision, to create.

Images